The USA Today research projects whose goal is to deliver, by 2010, software that can almost instantly translate Arabic and Mandarin Chinese with 90 to 95% accuracy. I’m no natural language researcher, but I’d be interested in knowing how they measure accuracy. I have seen some recent results on automated translation (French-English) and the results were alright some of the time, but even if you only screw up one word out of ten, is this good enough?
To anyone who has spent weeks crafting a ten pages paper: do you think you could get away by screwing up one word out of ten? Communication is a difficult task. It requires high accuracy most of the time.
Yes, I’m being hard. Automated translation is good and interesting work, but let us remain critical of the possibilities. While we lack strong AI, we cannot hope to replace human translators.